Research And Implementation Of Data Mining Mulder In Streaming Media Social Networking

With the development of the network technology, the data on the Internet is growingexplosively. It is becoming more and more difficult for the user to get accurateinformation. That is so-called Information Overload. The Streaming Media SocialNetworking also has the same problem. Also personalized recommendation system isthe way to solve this problem, there are still some problems in it. For example, sparsityproblem and cold start problem.We design the data mining component, which is composed by data mining moduleand recommendation module, to recommend friends and programs, based on theprogram of "Shenyang broadcast system for mobile phone". According to thecharacteristics of data, we design three independent recommendation algorithms.The main work is as follows:1. Various of recommendation algorithms have been researched. Arecommendation algorithm of items clustering and users clustering has been designed tosolve the problem of sparsity and time cost.2. Three independent recommendation algorithms have been designed accordingto the characteristics of data. The recommendation module is composed by the threealgorithms, based on friend’s relation, score and listening time.3. Similarity credibility and user attention have been proposed to improve thesimilarity formula.We have improved the recommendation systems according to the needs of thebroadcast system in this thesis. The achievement has also been tested by the analog data.The experiment proves that our improvement for the recommendation system bringsbetter results.